Quaternion-Based Attitude Estimation of an Aircraft Model Using Computer Vision

Area of Science:

• Aerospace Engineering
• Robotics and Control Systems
• Computer Vision

Background:

• Dynamic wind tunnel testing is crucial for aircraft flight dynamics research.
• Traditional methods may lack precision or require intrusive instrumentation.
• Accurate attitude estimation is vital for understanding aircraft behavior.

Purpose of the Study:

• To propose and validate a non-contact, off-board vision-based instrumentation method for dynamic wind tunnel testing.
• To enhance the accuracy of Euler angle identification using quaternions.
• To integrate vision-based data with on-board sensors using an Extended Kalman Filter.

Main Methods:

• Utilized Harris corner detection and Kanade-Lucas-Tomasi tracking for feature identification.
• Employed a stereo camera setup (side and top views) for 3D motion capture.
• Applied quaternion representation for attitude estimation and an Extended Kalman Filter for sensor fusion.

Main Results:

• The quaternion-based vision method achieved a 45.41% accuracy enhancement over the direction cosine matrix method.
• Root mean square errors for roll, pitch, and yaw rates were 0.0101, 0.0361, and 0.0036 rad/s, respectively.
• Integration with an Extended Kalman Filter further improved accuracy, achieving 98.61% for pitch rate estimation.

Conclusions:

• Quaternion-based attitude estimation offers superior accuracy for dynamic wind tunnel testing.
• The proposed vision-based method provides a robust, non-contact solution for aircraft dynamics analysis.
• Sensor fusion with an Extended Kalman Filter enhances the overall performance and reliability of the system.